Partition
On the Partition screen, you can check which nodes the database partitions of a SingleStore cluster are placed on and the replication status of each partition. You can also check whether table data is evenly distributed across partitions and whether any partition's row count is growing quickly, so you can find data skew and replication issues early.
Partitions and replication
SingleStore divides a database into multiple partitions and distributes them across Leaf nodes. For each partition, the original Master copy and a Replica copy for high availability are placed on different nodes. Reference tables, which keep the same data on every node, are marked separately as REF.
When partitions are evenly distributed, multiple nodes share the work. If data is concentrated on a specific partition or a Replica is missing, bottlenecks or failures can occur, so check placement and size together.
Query options
- Time: Set the query range. The default is the last 6 hours. Use quick select to choose the last 6 hours, the last 12 hours, Today, Yesterday, or 7 days ago. Real-time query is not supported.
- Cluster: Select the cluster to query. The first cluster in name order is selected initially.
Placement and size information is based on the values last collected before the end of the query range. Partition data is collected every hour. The top right of each panel shows the 1-hour cycle label and the last collection time.
Partition allocation
A grid shows which node each partition copy is on, by database. Rows are databases, columns are nodes, and each cell is a partition copy (p0, p1, and so on) or a Reference (REF).
- Partition N, Online M appears next to the database name, along with Offline N and No replication when applicable.
- The filled height of a cell is the size of the tables in that copy. The largest cell in the same database is 100%.
- Master copies have a solid border, and Replica copies have a dashed border and lighter color. Copies whose replication is not healthy are shown in the warning color.
| Legend | Meaning |
|---|---|
| Master | Original copy of the partition |
| Replica | Copy replicated from the Master |
| REF | Reference table that keeps the same data on every node |
| Replication anomaly | Copy whose replication status is not REPLICATING |
| Offline | Partition whose Master copy is not reported |
| No replication | Partition with no Replica. Normal in clusters that do not keep replicas |
Narrow the display range on the right side of the panel.
- Size: Select from All, ≥ 100 MiB, ≥ 1 GiB, Top 5, and Top 10 based on table size. The default is Top 5.
- Node: Select Leaf (default) or All nodes. Columns are listed in MA, CA, Leaf order.
Hover over a cell to see the partition name, Master or Replica, replication status, and the top 5 tables by size. Click a cell to open the partition detail drawer, or click a database name to open the database detail drawer.
Partition row increase
The chart shows partitions whose row count increased since the start of the query range, in order of increase. It plots the increase, not the total row count, so vertical axis values are in +N format. Partitions with no row increase are not shown, and row counts are based on the Master copy.
- On the right side of the panel, select the number of partitions to show from Top 5, Top 10, Top 15, and Top 20. The default is Top 5.
- Legend entries are in
database p<partition number> · +increase · multipleformat. The multiple is how many times the increase is relative to the average partition increase in that database. - Click a color box in the legend to hide or show the line. Click a legend name to open the Master detail drawer of that partition.
If the query range is shorter than 1 hour, only a single point appears. To see the growth trend, set the query range longer than 1 hour.
Table size
The table shows the size of each table and the size distribution across partitions. The default sort is Size in descending order. Narrow the display range with the Size filter on the right side of the panel.
| Column | Description |
|---|---|
Database | Database name |
Table | Table name |
Storage | Storage type (Rowstore, Columnstore). Reference tables marked (reference) |
Shard Key | Column used to divide data across partitions |
Size | Total table size |
Max Part Size | Size of the largest partition |
Rows | Total row count |
Compression | Compression savings (%). Orange under 33.3%, green at 66.7% or more |
Partitions | Bar chart of size per partition and reported partitions/total partitions |
The agent does not report partitions with fewer than 10,000 rows, so the number in the Partitions column can be lower than the total, such as 5/16.
- Partitions that are 1.5 times the reported partition average or larger are shown as orange bars.
- If unreported partitions cause skew, the bar chart colors stay the same and the number turns orange.
Hover over a bar to see the partition number, size and share, and Master node. Click a bar to open the partition detail drawer, or click a row to open the table detail drawer.
Partition detail drawer
Opens when you click a partition in a partition allocation cell, a bar in the Partitions bar chart, or the partition row increase legend. The title is in node › database › p<partition number> format.
| Area | Content |
|---|---|
| Overview | Master or Replica, Partition Type (Data, Reference), Total Partition Instances, nodes holding the Master and Replica copies, and Replication status |
| Table | Size and row count of each table in this partition compared with the table totals. Click a row to open the table detail drawer |
Database detail drawer
Opens when you click a database name in the partition allocation.
| Area | Content |
|---|---|
| Overview | Memory usage, disk usage, partition count, table count, and number of nodes in use |
| Table | Storage type, row count, memory, disk, and compression ratio of each table |
| Partition | Status, copy count, type, and Master and Replica nodes of each partition. Click a row to open the Master detail drawer of that partition |